UC Berkeley

Discussions about the measurement and conceptualization of implicit bias through the Implicit Association Test (IAT) have become increasingly prominent. Critiques center around the reliability of its measurement, such that individual level measurements tend to show a degree of instability that is not common for individual difference measures. The traditional scoring method of the IAT is the d-score, which uses reaction times to create a standardized difference score that serves as the measure of implicit bias. Sherman and colleagues’ (2005) quadruple process (QUAD) model argues that since the IAT is not a process-pure measure, both automatic and controlled processes should be estimated and analyzed. In the published literature the QUAD model has been validated at the condition (i.e. group) level. Reports of its use at the individual level (e.g., Wrzus et al. 2017; Calanchini et al. 2014; Gonsalkorale et al. 2014) suggests that the model can be fruitfully used at the individual level as well. We introduce such a methodology here, building upon the work of Sherman and Colleagues’ to validate the QUAD model parameters at the individual level. We offer an open-source, freely available R code for others to use. We were able to demonstrate that at the individual level, QUAD model parameters related to the IAT d-score in theoretically consistent ways while also providing novel information. We note that the d-score and iQUAD use very different scoring algorithms and suggest looking at both in unison to get the most information out of IAT data. Our findings provide a valuable tool to the field for disentangling, at the individual level, the d-score into its constituent components.